|Publication number||US7483986 B2|
|Application number||US 10/726,724|
|Publication date||Jan 27, 2009|
|Filing date||Dec 3, 2003|
|Priority date||Dec 3, 2003|
|Also published as||US20050125520|
|Publication number||10726724, 726724, US 7483986 B2, US 7483986B2, US-B2-7483986, US7483986 B2, US7483986B2|
|Inventors||James E. Hanson, Ian Nicholas Whalley|
|Original Assignee||International Business Machines Corporation|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (14), Non-Patent Citations (3), Referenced by (3), Classifications (9), Legal Events (4)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This invention relates generally to data processing systems and networks that include a plurality of autonomous software agents capable of self-organizing behavior resulting in the formation of inter-agent relationships and, more specifically, this invention relates to methods and apparatus for optimizing or tuning a topology of inter-agent relationships.
In multi-agent systems, agents form relationships among themselves according to their own internal preferences and decision-making logic. In general, there is no “dictator” function that determines the inter-relationships between agents. This level of self-organization is one of the defining features of multi-agent systems, and one of the most valuable.
Nevertheless, it is often desirable that the topology of the web of inter-agent relationships exhibit certain globally-defined properties. This is especially true if the multi-agent system is under the nominal administration of a single entity, such as a corporation, or was configured to solve some cooperative problem. For example, so-called “random” networks, such as the network of physical internet routers, in which the degree-distribution of nodes roughly follows a bell-curve, are relatively robust against targeted attacks on nodes, because the ill effects of the failure of any one node is approximately the same for all nodes. On the other hand, “scale-free” networks, in which the degree-distribution of nodes follows a power law, are relatively robust against random node failure, because most nodes are relatively unimportant to the overall network's functioning. Reference in this regard can be made to Albert, R., Jeong, H., and Barabasi, A.-L., “Error and attack tolerance in complex networks”, Nature 406:378-381 (2000).
For the purposes of this patent application, a “scale-free network” is any network of relationships in which the degree-distribution of the nodes is characterized by a power law. The “degree” of a node denotes the number of connections a given node has with other nodes, i.e., the number of relationships (or relationships of a given type) that a given node is in. Thus, for scale-free networks, the histogram of node-degrees has the feature that relatively fewer nodes are found to have relatively higher degrees, with the overall relationship following a power-law curve.
It is also typically the case that the overall desired network topology can change with time. For example, during typical functioning it is desirable for the system to maintain robustness against random failure; but in the event of a real or an expected attack, it would be even more desirable for the system to instead protect itself against the attack. For this to occur the system should exhibit an ability to continually adjust the statistical properties of the network topology.
A problem thus exists in the prior art as to how to dynamically and intelligently modify, or alter, or “tune” the topology of the network of relationships in an operating multi-agent system, in order to achieve a desired global behavior, such as robustness against attack or against random failure, without overly restricting the ability of agents to autonomously form and break relationships.
There is a body of established work on the relationship between network topology and system robustness, especially the tradeoff between robustness against attack versus robustness against random failure, as was noted above. In addition, however, there exists relevant work on the relationship between the degree of inter-connectivity among nodes in a localized neighborhood, and the rate at which chain reactions of failures are propagated. Reference in this regard can be made to Keeling, M., “The effects of local spatial structure on epidemiological invasions”, Proc. Roy. Soc. Lond. B 266, 859-869 (1999). This previous work indicates that changes to the topological structure of the network of relationships can have a profound effect on system dynamics, and in particular can result in improved system performance in the face of different kinds of failure or attack.
However, prior to this invention, there has been no description of a mechanism whereby these changes could be effected in the case of multi-agent systems.
There has also been extensive work on so-called “self-healing networks”, in which elements of a network (e.g., internet or telecommunications routers) respond to failures of nodes by reconfiguring themselves in a distributed fashion. These approaches are sometimes proactive, in the sense that they attempt to prepare the system to optimally respond to failures in advance of the failure actually occurring. Reference in this regard can be made to U.S. Pat. No. 6,061,735: “Network restoration plan regeneration responsive to network topology changes”, by James D. Rogers. However, this prior work does not address the need for systems to dynamically tune a preferred system topology to protect against different kinds of attack or failure, as the need arises.
The foregoing and other problems are overcome, and other advantages are realized, in accordance with the presently preferred embodiments of these teachings.
In one aspect this invention provides a data processing system that includes a plurality of self-organizing software agents capable of autonomously establishing relationships between themselves, where the totality of the relationships can be represented by a topological structure having nodes that each comprise one of the plurality of software agents and links between the nodes that comprise the established relationships. The data processing system further includes a topology tuning function that obtains information that is descriptive of the topology and, based at least in part on the obtained information and on at least one criterion, that makes one or more recommendations to at least one of the plurality of software agents that is intended to modify the topology. In one embodiment the topology tuning function is performed by a topology tuner software agent, while in another embodiment the topology tuning function forms a part of a system hosting function, such as a part of the operating system or other system software (e.g., an application server such as one known as the WebSphere™ Application Server that is available from the assignee of this patent application.)
In another aspect this invention provides, in a data processing system, a method to automatically tune a topology of relationships between a plurality of self-organizing agents. The method includes obtaining information that is descriptive of the topology and, based at least in part on the obtained information and on at least one criterion, making one or more recommendations to at least one of the plurality of agents that is intended to modify the topology of relationships between the plurality of self-organizing agents. This information may be obtained by the topology tuner agent by querying a system registry to determine identities of individual ones of the plurality of agents, and possibly also relationship information between the agents, and/or by querying individual ones of the plurality of agents to determine to which other agent or agents the individual one of the plurality of agents currently has a relationship.
In a further aspect this invention provides a topology tuner software agent operable in a data processing system that includes the plurality of self-organizing software agents. The topology tuner software agent includes computer program code for obtaining information that is descriptive of a topology of relationships between the self-organizing software agents from a system registry function or from the self-organizing software agents themselves; and also computer program code that is responsive to the obtained information and at least one criterion, for making at least one recommendation to at least one of the plurality of self-organizing software agents that is intended to modify the topology. Obtaining information can be performed periodically by topology tuner software agent to update the state of the topology, and the topology tuner software agent can make the recommendation when some topology threshold condition is satisfied.
In the various embodiments of this invention the at least one criterion may be a vulnerability of the topology to an attack directed to one or more of the self-organizing software agents, a vulnerability of the topology to a failure of one or more of the self-organizing software agents, and/or a rate at which the self-organizing software agents form new relationships.
The foregoing and other aspects of these teachings are made more evident in the following Detailed Description of the Preferred Embodiments, when read in conjunction with the attached Drawing Figures, wherein:
Upon starting, at Block 210 the topology tuner agent 20 first obtains information about the current topology of relationships in the system 10. This may take the form of an exhaustive list of all relationships between agents 40 in the system 10, possibly labeled with identifiers to distinguish between different types of relationship. This information can be conveniently represented as a graph in which the agents 40 are nodes and relationships are arcs connecting nodes. Alternatively, the topology information may be represented by statistical measures of system topology, such as the distribution of the number relationships the agents 40 have. For example, the topology tuner agent 20 may construct a histogram of the number of relationships the agents 40 are in. In such a histogram, the ordinate or x-coordinate is the number of relationships, and the abscissa or y-coordinate gives the number of agents found to be in that number of relationships. To determine whether the system topology was more scale-free or random in character, the topology tuner agent 20 compares the shape of the histogram against two alternative representations, the Gaussian distribution and the power-law distribution, using conventional statistical curve-fitting techniques. The closer the histogram is to a Gaussian distribution, the more “random” is the topology, while histograms that resemble the power-law distribution indicate a scale-free topology. A preferred method of gathering this topology information is shown in
Having gathered the topology information, at Block 220 the topology tuner agent 20 determines what type of tuning should be performed. This may be accomplished by starting with a high-level policy or goal, such as might be set by a system administrator, and calculating characteristics of the desired topology, and then comparing those characteristics with the characteristics of the measured topology. For example, the system administrator might specify that robustness under random node failure is a goal. The topology tuner agent 20 takes this predetermined goal and identifies a desired topology that most closely matches the topology required to achieve the goal, in this example a random graph. This implies that the topology tuner agent 20 attempts to achieve a normal or Gaussian distribution of connections per node (i.e., per each agent 40).
It should be noted that the practice of this invention does not require a general purpose mapping between high-level goals and topological characteristics. It is sufficient for this invention to have the mapping be even so simple as a set of hand-coded rules that apply to different predefined goals.
Once it has determined the type of tuning to be done, at Block 230 the topology tuner agent 20 contacts at least some of the agents 40 in the system 10 and directs them to modify their relationships in a particular way. This can be done in a multi-cast fashion, in which the topology tuner agent 20 sends the same message to all agents 40 at once; or it may be done in a more individualized, single-cast fashion, in which the agents 40 are contacted one by one, and each one potentially receives a different recommendation from the topology tuner agent 20.
After completing a tuning process or cycle, the topology tuner agent 20 determines at Block 240 whether re-tuning is necessary. Re-tuning is necessary in general to deal with two kinds of change, both of which may occur at any time: (a) the desired topology may change in such a way that the current topology is no longer desirable, or (b) the actual topology may change in such a way that it no longer has the desired characteristics. The former may occur, for example, if the system administrator changes the primary goal from “stability under random failure” to “security against future targeted attack”. The latter may occur, for example, because the agents 40 in the system 10 change their relationships, or due to a failure of one or more agents 40.
If the topology tuner agent 20 determines that re-tuning is necessary, control passes back to Block 220, otherwise the topology tuner agent 20 continues at Block 250 by monitoring the system topology and updating the topology information (possibly after a delay), and redetermining whether re-tuning is necessary. A preferred method of determining whether re-tuning is necessary and of updating the topology information is shown in
Therefore, in the preferred embodiment the topology tuner agent 20 contacts at Block 310 the directory service 50 and requests information about relationship registries in the system 10. If the directory service 50 returns the requested information at Block 320, the topology tuner agent 20 selects one or more of the registries (such as the Registry Service 30 shown in
After obtaining the topology information, either directly from the relationship register 30 or from the agents 40 themselves, the topology tuner agent 20 at Block 360 is enabled to calculate statistical measures of the system topology. For example, the topology tuner agent 20 may calculate the observed distribution in the number of relationships held by the agents 40. In the case where there are multiple types of relationships present in the system 10, the topology tuner agent 20 may calculate a distribution for each type of relationship.
There are various techniques for monitoring the topology of agent 40 relationships in the system 10. For example, the agents 40 may inform the topology tuner agent 20 whenever their relationships change; or the agents 40 may inform the relationship registry 30, which in turn informs the topology tuner agent 20, or the topology tuner agent 20 may poll the agents 40, or the relationship registry 30. Combinations of these techniques can be used as well.
At Block 420 the topology tuner agent 20 obtains, such as from the directory service 50, a list of agents 40 to make the recommendation to. This could be the entire population of agents 40, or a subset of the population for which a sub-graph topology is to be tuned. Alternatively, the list of agents 40 may comprise a subset of the population selected by relationship type, if the recommendation only applies to certain types of relationship.
Next the topology tuner agent 20 enters a loop to contact the agents on the list. If at Block 430 it finds that the list is empty, the topology tuner agent 20 exits the loop and this instance of the tuning process is complete (Block 470). Otherwise, the topology tuner agent 20 selects (and removes) an agent 40 from the list (Block 440). At Block 450 the topology tuner agent 20 may tailor its recommendation to the selected agent. For example, it may be the case that the selected agent must receive the recommendation in a different form than that of the general recommendation, it being assumed that this type of information may have been obtained along with the agent's directory entry. Another example concerns the case where the recommendation involves only certain types of relationship, and the selected agent is known not to be capable of forming relationships of that type. In this case the topology tuner agent 20 need not send the recommendation at all. At Block 460 the topology tuner agent 20 sends the customized recommendation and proceeds to select another agent 40 from the list.
It should be noted that the topology tuner agent 20 may measure or estimate the topology and/or distance from the desired topology characteristics at interim points while tuning, and thus may terminate the method early. In this case there would be another decision block in the loop, and an alternative exit to Block 470.
In general, there are various forms that the recommendations made by the topology tuner agent 20 can take. For example, the recommendations can include: (a) “re-form your relationships”, optionally with a relationship category; (b) “select an agent from a provided list”, optionally with a relationship category; and (c)“select an agent not on a provided list”, optionally with a relationship category. In this latter case, it can be appreciated that if the list includes one or more agents that the agent 40 currently has a relationship with, then the execution of this latter recommendation can have the effect of causing the agent 40 to break its relationship with one or more other agents 40 that are found on the list, and to form new relationships with other agents.
Depending on the particular situation, the topology tuner agent 20 can tune the network of agent 40 relationships for a number of different purposes. Some non-limiting examples are now presented.
The topology tuner agent 20 can tune the network of agent 40 relationships to enhance security against a future targeted attack (i.e., an attack that includes the choice of a target). It does this by requesting all or some of the agents 40 to re-form their relationships, without constraining the set of potential partners. The result is a shift toward a random-graph topology.
The topology tuner agent 20 can tune the network of agent 40 relationships to enhance robustness against random failure. It does this by requesting all or some of the agents 40 to re-form their relationships, giving preferential treatment to agents 40 in a selected list. The list is generated, and modified as the topology tuner agent 20 performs its assigned task, such that the resulting topology is or approaches that of a scale-free network.
The topology tuner agent 20 can tune the network of agent 40 relationships to provide protection against an attack on known or suspected target agent(s). It does this by requesting all or some of the agents 40 to re-form their relationships, and selecting partner agents that are not on a list of at-risk agents. The result is a network in which at-risk agents have few relationships, while the remainder of the agents 40 tend toward a random-graph topology.
The topology tuner agent 20 can tune the network of agent 40 relationships to provide protection against a possible failure of known agents. In a manner similar to the foregoing case, it does this by requesting all or some of the agents 40 to re-form their relationships, and selects partner agents that are not on a list of agents that are at-risk for failure. The result is a network in which at-risk for failure agents have few relationships, while the remainder of the agents 40 tend toward the random-graph topology.
The topology tuner agent 20 can tune the network of agent 40 relationships to provide for a reduction of “churn” or oscillations in system 10 topology, by reducing the rate of relationship change. This can be accomplished by selecting a subset of agents 40 (or all of the agents 40), and requesting those agents 40 in the selected subset to not re-form their relationships with other agents 40 for some specified period of time.
The topology tuner agent 20 can tune the network of agent 40 relationships to provide increasing system dynamism and responsiveness by increasing the rate of relationship change. This can be accomplished by selecting a subset of agents 40 (or all of the agents 40), and requesting those agents 40 in the selected subset to re-form their relationships with other agents 40. Depending on the frequency of the request, the agents 40 in the selected sub-set will tend to re-form their relationships with other agents at a rate that exceeds the rate of relationship reformation of the general population of the agents 40.
As was shown in
From the start state (Block 500), the topology tuner agent 20 first determines at Block 510 whether the topology information it has previously gathered is still sufficiently close to the desired topology. One simple technique to make this determination involves comparing a measured statistic, such as the distribution of number of relationships, with a goal statistic. The goal statistic may have been obtained in Block 220 of
If the measured topology does differ from the goal, then control transfers to Block 570 where the topology tuner agent 20 begins a new tuning cycle, also shown as the transition from Block 240 to Block 220 in
If the measured topology is still sufficiently close to the desired topology, the topology tuner agent 20 estimates whether the actual system 10 may have drifted sufficiently far from the desired topology in the time since the last measurement (Block 520). The actual method whereby the topology tuner agent 20 estimates the drift in the system 10 topology can be any suitable method, and can incorporate any standard technique for time-series estimation, such as linear forecasting.
If the estimated drift exceeds an acceptable threshold limit, the topology tuner agent 20 begins another tuning cycle by transferring to Block 540, otherwise the topology tuner agent 20 waits at Block 530 for some period of time. Note that, like the desired topology, the acceptable threshold limit may have changed since the last comparison. For example, the system administrator may adjust the threshold to reflect changing high-level goals, or the topology tuner agent may calculate the threshold dynamically. Thus, the threshold that determines when to begin a new topology tuning cycle may be a variable threshold. The period of time may follow a predetermined schedule, or it may be dependent on the estimated drift rate, or it may be dependent on some other parameter or parameters (such as a historical drift rate under some certain set of system 10 conditions). The waiting period at Block 530 may be interrupted if the topology tuner agent 20 receives a new topology goal, or if it receives an express notification that the topology has changed.
At Block 540 the topology tuner agent 20 updates its measured topology information, which corresponds to Block 250 of
Following re-measurement of the system 10 topology, at Block 550 the topology tuner agent 20 updates its model of the system drift, e.g., by adjusting the estimated linear drift rate, and then returns to Block 510.
It should thus be appreciated that the decision as to when to begin another tuning cycle (
Based on the foregoing description of the presently preferred embodiments of this invention, it should be appreciated that an aspect of this invention is to execute, in the multi-agent system 10, the functionality inherent in the topology tuner agent 20. The topology tuner agent 20 gathers statistical information about the relationship-network's topology (such as the degree-distribution of nodes), and then makes recommendations to individual ones of the agents 40 to modify their relationships (if needed).
The topology tuner agent 20 is preferably endowed with special privileges, giving it sufficient nominal authority to make its recommendations to the other agents 40 effective, but otherwise preferably behaves as a “peer” agent. For example, the topology tuner agent 20 preferably uses the same messaging infrastructure as all other agents 40.
An aspect of this invention is the operation of the topology tuner agent 20 in gathering information about the network topology of the system 10, as was described above in relation to Block 210 of
As was noted above, it is also within the scope of this invention to classify relationships between agents 40 according to some classification scheme (e.g., “relationships concerning storage services” versus “relationships concerning bandwidth services”), thereby identifying different kinds of links between topological network nodes (i.e., between agents 40). To perform this function the topology tuner agent 20 need not be endowed with any “understanding” of the meaning of the various relationship categories, but only with the ability to identify to which category a given link belongs. As such, the category determination may be based simply on keyword matching.
Once the network topology graph is constructed, the topology tuner agent 20 identifies those agents 40 (i.e., network nodes) with an undesirable property or properties. For example, to make the network more robust against attack, the topology tuner agent 20 locates the nodes with the greatest number of links (of one or more categories) with other nodes. It then finds those “neighbor” agents (i.e., the agents linked to the agent that has been identified), and recommends to one or more of those agents that they link to a different node (i.e., that they link to a different one of the agents 40).
This recommendation may take several forms, from expressly identifying a particular one of the agents 40 to link to, to identifying a set of preferred agents 40 to link to, or to simply requesting that the agent break the existing link and select any other agent to link to according to its own preferences.
In a preferred embodiment of this invention the topology tuner agent 20 also requests that each neighbor agent return, to the topology tuner agent 20, a ranked list of alternate agents that the neighbor agent would find acceptable. Using this returned list, the topology tuner agent 20 is enabled to select one or more other agents that a particular agent 40 would find preferable. However, the topology tuner agent 20 is not obligated to offer a particular agent 40 one of its preferred agents to form a relationship with, as the topology tuner agent 20 may have knowledge that the preferred agent 40 is actually one that is a probable target for an attack, or a probable candidate for a failure.
An aspect of this invention involves a consideration of, and a compensation for, system 10 dynamics. During the time that the topology tuner agent 20 is gathering information and making recommendations, the agents 40 in the system 10 are carrying out their assigned tasks. This implies that the network of relationships is most likely continually changing, independent of the activities of the topology tuner agent 20. It is thus preferred that the topology tuner agent 20 incorporate the dynamics of the system 10 into its internal model.
This can be accomplished in several ways. For example, and as was discussed above in reference to
The foregoing description has provided by way of exemplary and non-limiting examples a full and informative description of the best method and apparatus presently contemplated by the inventors for carrying out the invention. However, various modifications and adaptations may become apparent to those skilled in the relevant arts in view of the foregoing description, when read in conjunction with the accompanying drawings and the appended claims.
As but some examples of modifications and adaptions that may occur to those skilled in the art, it should be noted that the topology tuning function need not be implemented as the software agent 20, but could, instead, be implemented as a part of the hosting environment, such as by an entity that resides in the hosting environment.
Also, the foregoing description of the topology tuner agent 20 has assumed that there is no centralized control of the system 10, and possibly no centrally stored information that is descriptive of the topological properties of the agent 40 relationships. However, the topology tuner agent 20 may also be of use in situations where there is some degree of centralized control and/or centrally stored information.
For example, it may be the case that all of the agents 40 in a multi-agent system 10 have been programmed to follow a common set of policies. Typically in such situations, these policies specify a core set of common behaviors to which the agents 40 must conform, but do not fully specify the behavior of each individual one of the agents 40. In this case the agents 40 can obtain updates to the common set of policies from time to time, and change their behavior accordingly. In such a situation the recommendations of the topology tuner agent 20 may be incorporated into the common set of core policies and transmitted to the other agents 40 by means of a policy update, rather than by direct interaction between the topology tuner agent 20 and the agents 40 of the system 10.
The use of other similar or equivalent techniques and procedures may thus be attempted by those skilled in the art. However, all such and similar modifications of the teachings of this invention will still fall within the scope of this invention.
Furthermore, some of the features of the present invention could be used to advantage without the corresponding use of other features. As such, the foregoing description should be considered as merely illustrative of the principles of the present invention, and not in limitation thereof.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US5091920 *||Oct 31, 1990||Feb 25, 1992||Fujitsu Limited||Threshold valve control system in a receiver circuit|
|US6061735||Oct 23, 1997||May 9, 2000||Mci Communications Corporation||Network restoration plan regeneration responsive to network topology changes|
|US6697325 *||Dec 7, 1999||Feb 24, 2004||Nortel Networks Limited||System, device, and method for expediting reconvergence in a communication network|
|US6882627 *||Jun 14, 2001||Apr 19, 2005||Tropic Networks||Methods and apparatus for selecting multiple paths taking into account shared risk|
|US7013395 *||Mar 13, 2001||Mar 14, 2006||Sandra Corporation||Method and tool for network vulnerability analysis|
|US7031321 *||Jun 10, 2002||Apr 18, 2006||Koninklijke Philips Electronics N.V.||Dynamic network and routing method for a dynamic network|
|US7184421 *||Dec 21, 2001||Feb 27, 2007||Itt Manufacturing Enterprises, Inc.||Method and apparatus for on demand multicast and unicast using controlled flood multicast communications|
|US20030046583 *||Aug 30, 2001||Mar 6, 2003||Honeywell International Inc.||Automated configuration of security software suites|
|US20040044891 *||Sep 4, 2002||Mar 4, 2004||Secure Computing Corporation||System and method for secure group communications|
|US20040059812 *||Dec 14, 2001||Mar 25, 2004||Shmuel Assa||Topology information system for a managed world|
|US20040081166 *||May 1, 2002||Apr 29, 2004||Stanforth Peter J.||System and method for using an ad-hoc routing algorithm based on activity detection in an ad-hoc network|
|US20040117624 *||Sep 12, 2003||Jun 17, 2004||Brandt David D.||System and methodology providing automation security analysis, validation, and learning in an industrial controller environment|
|US20040218582 *||Apr 30, 2003||Nov 4, 2004||Harris Corporation||Predictive route maintenance in a mobile ad hoc network|
|US20050080894 *||Oct 9, 2003||Apr 14, 2005||John Apostolopoulos||Method and system for topology adaptation to support communication in a communicative environment|
|1||*||C. Bettstetter. On The Minimum Node Degree and Connectivity of a Wireless Multihop Network. Technische Universitat Munchen. MOBIHOC'02. Jun. 9-11, 2002.|
|2||M. J. Keeling, The Effects of Local Spatial Structure on Epidemiological Invasions, The Royal Society, pp. 859-867.|
|3||Réka Albert, Hawoong Jeong and Albert-László Barbási, Error and Attach Tolerance of Complex Networks, Nature: vol. 406, Jul. 27, 2000, pp. 378-381.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US9043920||Oct 17, 2012||May 26, 2015||Tenable Network Security, Inc.||System and method for identifying exploitable weak points in a network|
|US9088606||Dec 3, 2012||Jul 21, 2015||Tenable Network Security, Inc.||System and method for strategic anti-malware monitoring|
|US20100057902 *||Aug 28, 2008||Mar 4, 2010||Novell, Inc.||Scal-free management networks|
|International Classification||G06F15/173, H04L12/24|
|Cooperative Classification||H04L41/042, H04L41/046, H04L41/12|
|European Classification||H04L41/04A, H04L41/04C, H04L41/12|
|Dec 3, 2003||AS||Assignment|
Owner name: INTERNATIONAL BUSINESS MACHINES CORPORATION, NEW Y
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:HANSON, JAMES E.;WHALLEY, IAN NICHOLAS;REEL/FRAME:014778/0534
Effective date: 20031202
|Sep 10, 2012||REMI||Maintenance fee reminder mailed|
|Jan 17, 2013||SULP||Surcharge for late payment|
|Jan 17, 2013||FPAY||Fee payment|
Year of fee payment: 4